fix: address CodeRabbit review feedback for SPEC-21

fix: address SPEC-21 code quality issues

refactor: implement proper middleware architecture (SPEC-21)

fix: PubMed JSON parsing (SPEC-20)

fix(tests): Address CodeRabbit review feedback

fix(orchestrator): P2 - Silence ExecutorCompletedEvent UI noise

fix(types): Ignore CompiledStateGraph type-arg for cross-version compat

fix(deps): Update urllib3 to 2.6.0 for security fixes

style: Address CodeRabbit nitpicks

refactor(orchestrator): implement semantic progress tracking

fix(deps): SPEC-18 Agent Framework Core Upgrade to 1.0.0b251204 (#131)

refactor(cleanup): Remove Anthropic + Modal partial wiring (P3 Tech Debt) (#130)

refactor(orchestrator): Introduce WorkflowState dataclass (Priority 7) (#129)

refactor(factory): Implement Provider Registry Pattern (Priority 6) (#128)

refactor(prompts): Unify prompt storage in src/prompts/ (Priority 5) (#127)

refactor(config): Remove dead config and Anthropic references (Priority 4)

refactor(orchestrator): DRY synthesis + constants + clean imports

fix(orchestrator): Force synthesis when ReportAgent doesn't run (P1)

fix(p2): Resolve duplicate report content and first-turn timeout bugs (#123)

fix(huggingface): P1 Free Tier tool execution - Remove premature marker (#121)

fix(gradio): Prevent examples from auto-submitting on click

fix(coderabbit): Address all CodeRabbit review findings

fix(arch): End-to-end BYOK support for unified architecture

fix(P1): Replace stale 72B model fallbacks with 7B to avoid Novita 500 errors

fix(P0): Auto-detect OpenAI provider from BYOK api_key prefix

fix(P1): Switch to Qwen2.5-7B to avoid HuggingFace third-party provider failures (#118)

fix(P0): Implement Accumulator Pattern to resolve Repr Bug (#117)

fix: Address CodeRabbit review feedback

fix(P0): Complete HuggingFace tool calling integration and document framework display bug

fix(P0): Add function invoking marker and improve ChatMessage extraction

fix(P0): Complete bi-directional tool support for HuggingFace (sync+streaming)

fix(P0): Implement proper AIFunction serialization for HuggingFace

fix(P1): Resolve HuggingFace 401 - invalid token was root cause

fix(auth): Robust HF_TOKEN loading and debug logging for Partner API auth failure

fix(P1): Switch from Llama-3.1-70B to Qwen2.5-72B for HuggingFace free tier

feat(SPEC-16): Unified Chat Client Architecture (#115)

fix: Address all CodeRabbit review nitpicks

fix(perf): Implement P2 Phases 2 & 3 (Pre-warming + Gradio Progress)

fix(advanced): P2 dead zones granular init events

chore: Address CodeRabbit review feedback

fix: P1 Advanced Mode chain-of-thought interpretability (#106)

docs: Clarify config parameter behavior in factory.py

fix: P0 Advanced Mode timeout synthesis + CodeRabbit recommendations

feat: Implement Free Tier synthesis using HuggingFace Inference

fix: P0 provider mismatch and code quality audit fixes (#102)

feat: SPEC_15 Advanced Mode Performance Optimization (#101)

feat(clinicaltrials): SPEC_14 Outcome Measures (#100)

feat(search): SPEC_13 Evidence Deduplication (#98)

fix: use logger.error with stack trace for synthesis failures (CodeRabbit)

fix(ui): CSS autofill override and text centering